Machine learning-assisted classification of carbon nanotube, Polymer Matrix, and graphene reinforced nanocomposites using multivariate material characterization data

Abstract Composite materials reinforced with nanomaterials such as carbon nanotubes (CNTs) and graphene have attracted considerable attention owing to their superior mechanical strength, stiffness, lightweight nature, and enhanced thermal properties, making them suitable for aerospace, automotive, renewable energy, marine, and biomedical applications. Conventional characterization techniques such as ultrasonic testing and thermal analysis methods remain essential for evaluating microstructural features, crystallographic behavior, defect formation, thermal stability, and thermophysical performance of composite materials. However, these experimental approaches are often time-consuming and less suitable for rapid large-scale material screening and classification. In this study, machine learning (ML) techniques are employed to develop an automated framework for composite material component classification using multivariate material property data relevant to structural and thermal characterization. Multiple ML classifiers were evaluated, and the XGBoost model achieved the top classification accuracy as 94.7% using 10-fold cross-validation, with an F1-score of 95%. To improve predictive reliability and computational efficiency, a hybrid hyperparameter optimization strategy combining Randomized Search for global exploration and Grid Search for local refinement was implemented. Confusion matrix analysis demonstrated consistently high precision, recall, and F1-scores across the investigated classifiers, confirming robust predictive capability. The findings highlight the rise of artificial intelligence in materials science, particularly for accelerating material screening, thermal-performance assessment, and data-driven composite material analysis.